Machine learning based adaptive congestion window adjustment for Congestion Aware Routing in Cross Layer Approach Handling of Wireless Mesh Network

Abstract

Among different researches on Wireless Mesh Networks (WMN), the Cross-layer Handling Link Asymmetry (CHLA) scheme has been enhanced with QoS-based Congestion Avoidance using Congestion Aware Routing (QSCACAR) technique. In this approach, QoS requirements were achieved by estimation of signal strength, network capacity, MAC scheduling, link scheduling and slot assignment. In addition, congestion was controlled based on the bandwidth management mechanism which is according to the congestion window on the dynamic traffic conditions. Based on these approaches, congestion over the network was controlled. However, the congestion window size was not precisely adjusted and selected by using these approaches. Therefore in this article, CHLA-QSCACAR is enhanced by including adaptive congestion window size based on the machine learning algorithm (CHLA-MQSCACAR). In this approach, congestion window size is predicted for adjusting the congestion in the next transmission. The Support Vector Machine (SVM) algorithm utilizes the congestion balance status as label and congestion control parameters as input attributes. Initially, congestion balance status of different transmissions and their corresponding input congestion control parameters are gathered. The collected database is trained with aid of SVM classifier. Then by using the trained SVM model, the congestion status of current transmission with congestion window size is predicted. The predicted congestion window is utilized for achieving congestion balance for the consecutive transmissions. Finally, the experimental results show that the performance effectiveness of the proposed CHLA-MQSCACAR compared to the existing CHLA-QSCACAR method in terms of different metrics such as routing overhead, bit error rate, end-to-end delay, throughput and latency.